ChatGPT parent OpenAI’s anti-cheating technology has been ready for about a year|Marketcomlabo|CC BY-SA 4.0
ChatGPT maker OpenAI has an AI writing detection tool that is 99.9% effective but hasn’t released it due to internal debates, reports the Wall Street Journal.
The anti-cheating technology has been ready for about a year. The decision to release the tool has been complicated by concerns about user reactions and the potential for the tool to affect ChatGPT’s performance.
The company found in a survey of loyal CatGPT users that nearly one-third of them would be turned off by the anti-cheating tool.
Additionally, to make the tool work, the AI system powering the ChatGPT bot would have to leave watermarks—only detectable by the technology—in all the generated content.
Some educators are eager for such a tool to help combat AI-assisted cheating, as AI-generated content in schools is rising.
However, OpenAI is concerned about the possibility of the watermark being easily bypassed by hackers and rival companies. Executives, including CEO Sam Altman, have discussed the technology extensively but have yet to finalize a decision.
Not the only one, Google has an AI detection tool called SynthID to detect text generated by its Gemini AI bot. It is in beta testing and not widely available.